Sports Video Summarization Using the Deep Learning Techniques

As sports content rapidly expands, fans struggle to keep up with live action using traditional highlight methods, which often miss key moments or include irrelevant segments. This paper introduces a hybrid sports video summarization system combining the YOLO deep learning framework with traditional video processing to detect critical events (e.g., goals, fouls, sixes) in real time. YOLO achieves <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$85-90 \%$</tex> accuracy in detecting players, the ball, and scoreboards, while traditional techniques improve robustness in lowresolution or dynamic scenes (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$15-20 \%$</tex> fewer missed detections). The system generates concise video highlights and text summaries using NLP, reducing summarization time by 35-40% compared to manual methods. Evaluated on football, cricket, and basketball datasets, the hybrid approach achieves 95 % F1-score in cricket six detection, 92 % precision in football goal identification, and 90 % relevance ratings for basketball dunk highlights. Fans can access highlights 4 x faster than full-match replays, with text summaries achieving a BLEU score of 0.82 against human annotations. By merging YOLO's speed with classical video analysis, this system delivers efficient, context-aware sports summarization, enabling fans to stay updated with 95 % key-moment coverage.